Crypto Sentiment Dashboard

A Streamlit dashboard for analyzing crypto news sentiment (BTC / ETH / SOL / OTHER) using VADER.
It includes interactive charts (Plotly), export-ready PNGs (RGB-safe), and a print-quality PDF report (ReportLab).

Disclaimer: This project is for research and monitoring purposes only. It is not financial advice.

---

✨ Features

• Interactive charts — Source & label bar charts, daily time-series, and correlation heatmap.
• Crypto domain insights — Auto-generated summary, bullet-point analysis, recommendations, and conclusion.
• Export-ready images — Large-canvas PNGs with generous margins, forced RGB to avoid disappearing lines in PDFs.
• One-click PDF — A neatly paginated report with all charts and summary sections.
• Powerful filters — Label, source, and UTC date range.
• Pandas compatibility — Hides table index with a safe fallback for older pandas versions.

---

🚀 Quick start

# 1) (optional) create a virtual environment
python -m venv .venv
# activate it:
#   Windows: .venv\Scripts\activate
#   macOS/Linux: source .venv/bin/activate

# 2) install dependencies
pip install -r requirements.txt

# 3) run the app
streamlit run streamlit_app.py

The app reads data from data/news_raw.csv.

---

📦 Requirements

• Python 3.10+ (recommended)
• streamlit, pandas, numpy
• plotly, matplotlib
• kaleido (PNG export backend for Plotly)
• reportlab, Pillow (PDF builder and image backend)

All pinned in requirements.txt.

---

🗂️ Data schema

The loader (sentiment.py) normalizes your columns automatically. Minimum required:

• publishedAt / published / date → normalized to publishedat (UTC timezone)
• source — outlet name/domain
• compound — VADER score in [-1, +1]

Optional:

• channel, title, description, url, label (if label is missing, it is inferred from text: BTC/ETH/SOL, else OTHER)

Example row:

publishedAt,source,compound,title,url,label
2025-08-01T10:30:00Z,CoinDesk,0.21,"BTC breaks range","https://example.com/article",BTC

Tip: If your dataset is large or sensitive, keep it out of your repo and list data/*.csv in .gitignore.

---

📊 Charts & PDF exports

When you click Build PDF report in the UI, the app:

1. Saves high-resolution PNGs for each chart with large margins so labels never get cropped.
2. Forces RGB (removes alpha) on the PNGs to prevent thin lines from disappearing in PDFs.
3. Assembles a polished, paginated PDF using ReportLab — one chart per page, plus summaries.

About the time-series edges
The PDF version uses a small time padding by default to avoid edge clipping. You can switch to fully edge-to-edge by editing the timeseries_by_label(..., theme_for_pdf=True) x-axis range in streamlit_app.py.

---

🧱 Project structure

.
├─ streamlit_app.py        # Streamlit UI (Plotly + PDF trigger)
├─ sentiment.py            # Data loader, domain analysis, print-safe charts (matplotlib), PDF builder
├─ requirements.txt
├─ data/
│  └─ news_raw.csv         # Your dataset (kept local by default)
├─ charts/                 # Generated PNGs (auto-created)
└─ LICENSE                 # MIT © 2025–present wiqilee

---

⚙️ Configuration

Adjust label colors/order in both modules if desired:

LABEL_COLORS = {"BTC": "#22c55e", "ETH": "#ffffff", "SOL": "#fde047", "OTHER": "#94a3b8"}
LABEL_ORDER  = ["BTC", "ETH", "SOL", "OTHER"]

---

🛠️ Troubleshooting

• kaleido export error
  Ensure kaleido==0.2.1 is installed:

  pip install --upgrade kaleido

• Time-series lines missing in PDFs
  Already handled by forcing RGB on exported PNGs. Make sure Pillow is installed.

• Table index won’t hide (old pandas)
  The app falls back from styler.hide_index() to styler.hide(axis="index") automatically.

---

🗺️ Roadmap

• Automated news ingestion
• Alternative sentiment models beyond VADER
• One-click deploy to Streamlit Community Cloud

---

🤝 Contributing

Contributions are welcome! Please open an issue or submit a pull request.
Make sure your PR includes a clear description and, when possible, screenshots of UI changes.

---

📜 License

MIT © 2025–present wiqilee — see LICENSE for details.